[Bug c++/46706] New: global constructors does not work properly since changes from bug 29141

["mschulze at ivs dot cs.ovgu.de" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46706

           Summary: global constructors does not work properly since
                    changes from bug 29141
           Product: gcc
           Version: 4.4.0
            Status: UNCONFIRMED
          Severity: critical
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: mschulze@ivs.cs.ovgu.de
                CC: mschulze@ivs.cs.ovgu.de
            Target: avr


In my opinion the bug 29141 is still pending and not fixed. I reported it there
but until now no reaction and I am not able to reopen the bug thus I open a new
one.

The proposed fix of bug 29141 rises another issue. Since this change, global
constructors does not work properly anymore. The fix uses register r20 within
the __do_global_ctors loop but this register could be clobbered by a
constructor that is called, leading to destructing the exit condition of the
__do_global_ctors loop. In new version of gcc (see bug 45263 but still not
confirmed) this is weakly fixed by pushing and poping r20 around the
constructor call. I would suggest using a register between r2-r17 instead of
r20. According to the compiler abi this should be a register that the called
routine has to save if it needs to use it.

In my eyes, this is a critical or maybe a blocker bug, because global objects
can not be used as long as r20 is used unprotected without pushing/poping, or
by using another register instead that is not faced by this problem. 

Affected gcc versions 4.4.0 - 4.5.1